كيف يعرف المساعد الصوتي أنك انتهيت من الكلام؟
كيف يكتشف المساعد الصوتي أن المتحدث أنهى دوره؟ شرح لطرق اكتشاف نهاية الكلام، من الصمت وVAD إلى النماذج التي تفهم النص والصوت معاً.
til
مقالات وتجارب وملاحظات حول til.
نموذج صقر بحجم 50 مليون لتوليد البيانات بالعربي يعمل علي CPU
ماهو نموذج Sakr صقر المخصص لتوليد البيانات باللغة العربية ومدرب علي حوالي 66 مليون صفحة من fine-web dataset يعمل بسرعه فائقة ونتائج متوسطة. شرح تفصيلي.
ماهي Likve Kit مقدمة سريعة
هي مشروع مفتوح المصدر لبناء وكلاء ذكاء إصطناعي تتعامل مع الصوت و الصورة بشكل لحظي وتشمل كل ما يحتاجه المبرمج
فوائد و أضرار عمل قناة يوتيوب لمشاركة المعرفة
بشارك خواطر تجربة إنشاء قناة لمشاركة المعارف في تخصصي وهو الذكاء الإصطناعي والبرمجة في عام ملئ بالتشتت والسطحية في نشر المعرفة.
Using Dinov3 Search to Identify Fonts in Historical Arabic Books
Building OCR for historical Arabic manuscripts is hard when you don't know what fonts the text resembles. I built a simple image similarity pipeline using DINOv3 and Qdrant to match page scans against 300 Arabic fonts — and found one font d
Latest TTS Architecture and differences | 2026
Gallery of Popular TTS Architecture and Papers, Pros and Cons and unique ideas from them, quickly explained.
Kareem CV of Failures 'Rejections build Career'
A quick summary of My personal failure and why you should write also a cv of failures to help you track the problems you create over time
Hacker AI Engineer Interview 2026 Example
HackerRank AI engineer interview review 2026. prompt engineering, SQL, and RAG coding challenges that test memorization over real-world AI skills.
Vector Databases - O'reilly By Nitin Borwankar
Review of a new book about Vector database from O'reilly
HyperRun + ColGrep: A Self-Hosted Alternative to RunLLM
HyperRun is an open-source, self-hosted alternative to RunLLM. Add an AI-powered "Ask AI" chat widget to any docs site using ColGrep semantic code search, Lisette LLM orchestration, and FastHTML. Supports BYOK, multi-provider, and embeds wi
HyperRun Deep Dive: Streaming Chat with FastHTML, HTMX, and SSE
How I built a real-time AI chat widget with zero React, zero WebSockets, and almost zero custom JavaScript. A deep dive into FastHTML's server-rendered approach — SSE streaming, DaisyUI modals with checkbox toggles, OOB swaps, session-based
Building SEO RAT: A Free, Open-Source SEO Tool
Why I'm building SEO RAT — an open-source SEO tool focused on real gains from Google Search Console, smart visualizations, and AI-powered keyword analysis.
OpenCode: Why It's Different and How to Use It with Python
Learn how OpenCode works as programmable AI coding infrastructure. Built in TypeScript, controlled with Python — covering agents, skills, PLAN.md, and MLflow integration.
Super Productivity Review: Best Free Todo App (2026)
A detailed review of Super Productivity, the free open source todo app. Explore how I use it for managing AI research, development tasks, and daily productivity.
Diversifying Search Results with Pyversity and Qdrant
Learn how to diversify search results in a RAG system using Qdrant and Pyversity. Covers MMR, DPP, and other diversity algorithms with practical Python examples.
AI Research Challenges: The GPU Barrier for Researchers
Exploring the challenges of AI research without high-end compute, focusing on feedback loops and the need for hardware like the RTX 5090.
Rust for ML: Building AI Tools and Data Pipelines in 2026
Learning Rust for ML engineering—building AI tools with Polars, Qdrant, and planning semantic search projects like Pyversity and Semhash
Arabic NLP Fundamentals: Sparse Embeddings, BM25 & Dynamic Programming
Rebuilding NLP fundamentals for Arabic: TF-IDF, BM25, sparse embeddings, minimum edit distance, and learning problem-solving through LeetCode and Jurafsky's Speech and Language Processing book.
AraModernBert vs AraBERT v2: Arabic Tokenizer Comparison
AraModernBert vs AraBERT v2 tokenizers for Arabic NLP. Test results: AraModernBert is 15-71% more efficient. Fertility, compression ratio, and practical performance on MSA and dialect text.
AI Engineering Careers: Connecting with Minds in Machine Learning
Reflections on landing an AI engineering role, the importance of professional connections, and following the right experts in the machine learning field.
PyLate Tutorial: Vector Indexing and Retrieval with ColBERT
PyLate tutorial for ColBERT vector retrieval. Learn dataset formats, PLAID indexing, evaluation metrics, and practical code examples for dense retrieval systems.
Meta's RAPTOR RAG, DFloat11 Compression & Pyrefly Guide
A complete guide to Meta's RAPTOR RAG technique, DFloat11 lossless model compression, and the Pyrefly Python type checker with implementation details.
Does Brave Leo Run Locally? Privacy-First AI with Ollama and BYOM
Yes, Brave Leo runs locally with anonymous requests and local chat storage. Configure Brave Leo to use any local LLM via BYOM: Ollama, vLLM, or SGLang for private AI in your browser.
البوصلة الحقيقية: تأملات في تزكية النفس وإخلاص النية في طلب العلم
تأملات حول التوجه إلى الله وتزكية النفس، والتحديات الإيمانية التي تواجه الباحث في مجال الذكاء الاصطناعي للتحرر من الأنانية وتحقيق الإخلاص في طلب العلم.
How to Connect Porkbun Domains to Vercel and Cloudflare (Step-by-Step)
Connect Porkbun domains to Vercel via DNS records and Cloudflare for Railway. Complete steps for domain transfers, WHOIS privacy setup, and DNS propagation troubleshooting.
AI Research Blogging: Why Substack Matters for Researchers
Insights on starting a Substack for AI researchers. Learn how blogging improves networking, writing skills, and knowledge management for AI engineering.
Explore the Qdrant Blog core concepts | Part 1
This a 2 hours daily exploring and summarizing Qdrant Blogs part 1
Learning in Public: Why I Started My TIL Journey
Starting a Today I Learned (TIL) series to overcome perfectionism in technical writing and build professional connections within the machine learning community.